엔비디아 대 브로드컴: 향후 5년간 더 나은 AI 반도체주는 어디인가?

Nvidia (NVDA -2.94%) and Broadcom (AVGO -4.35%) are two of the biggest names in AI investing. Both provide a massive amount of the computing power that AI companies need, and they're growing rapidly as a result. But which one is the better buy today -- the graphics processing unit giant or the custom chip leader? Nvidia and Broadcom are approaching computing from different directions Nvidia and Broadcom are both capturing large slices of the AI spending pie, but in different ways. Nvidia's graphics processing units (GPUs) have been the industry standard AI computing units since the AI race kicked off, but it's not resting on its laurels. It recently started shipping its latest generation AI architecture, Vera Rubin. This platform, built around Vera central processing units (CPUs) and Rubin GPUs, is far more advanced than the previous Blackwell platform. When tasked with inference, it reduced token cost by a factor of 10, and for training, it cuts the number of GPUs required for a given workload by 75%. Those are huge gains in power. GPUs are great for a wide variety of computationally heavy tasks, and Nvidia clients can use its popular CUDA platform to program them for those workloads, but that flexibility comes with higher upfront costs. Broadcom takes a different approach: It makes custom application-specific integrated circuits (ASICs) that are suitable only for narrowly defined workloads. It has an impressive client list that includes the likes of hyperscalers Alphabet and Meta Platforms, as well as AI labs such as Anthropic and OpenAI. That's a who's who in the AI realm, and more clients could join the list when they see what a custom AI chip can do. These computing units can outperform GPUs in both performance and cost when running a properly configured workload that suits their design. However, they would fall flat if tasked with a workload outside of their core competency. NASDAQ: AVGO Key Data Points There isn't much of a debate as to which company has better technology, as it's not really a fair comparison. These two provide completely different products, though they are used for the same purposes. The question is whether AI hyperscalers will continue to use broad-purpose computing units like GPUs, or switch to favor specialized ASICs like the ones Broadcom designs. I think the future will be a mixture of both, but may lean more heavily toward ASICs. This gives Broadcom a slight edge, and that's reflected in both companies' guidance. Broadcom is expected to grow faster than Nvidia In its upcoming fiscal year, Nvidia expects to grow revenue at an impressive 70% rate. That's quite rapid given its size, and investors shouldn't be disappointed by that guidance. Furthermore, Nvidia has a strong track record of outperforming its own forecasts, so it may do even better than that. NASDAQ: NVDA Key Data Points Broadcom expects even more rapid AI semiconductor growth next year. It informed investors that it should double its AI semiconductor revenue to $115 billion in 2027, and then double it again to $230 billion in 2028. That's notably faster growth than Nvidia is anticipating for itself. Where might those trajectories put these stocks five years from now? Over the next five years, I'd expect major AI data center growth, as we're a long way from having the computing power infrastructure that would be necessary to power an AI-first economy and society. The ongoing build-out of that infrastructure will provide huge growth opportunities for both companies, but with workloads starting to become more established, I think Broadcom will have the growth advantage. That's not to say Nvidia won't do well. There are plenty of things that Nvidia processors can do that Broadcom's cannot do. For investors looking for a smart AI play, I think splitting an investment between these two companies would make a ton of sense. They're both bound to be among the biggest beneficiaries of the AI trend over the next five years.